Monthly Archives: October 2024

To what extent is science a strong-link problem?

Here’s a fascinating and worrying news story in Science: a top US researcher apparently falsified a lot of images (at least) in papers that helped get experimental drugs on the market — papers that were published in top journals for years, and whose problems have only recently become apparent because of amateur sleuthing through PubPeer.

I’m going to wane philosophical for a minute. In general I’m very sympathetic to Adam Mastroianni’s line “don’t worry about the flood of crap that will result if we let everyone publish, publishing is already a flood of crap, but science is a strong-link problem so the good stuff rises to the top”. I certainly don’t think we need stronger pre-publication review or any more barrier guardians (although I have reluctantly concluded that having some is useful). But when fraudulent stuff like this does in fact rise to the top in what seems to be a strong-link network — lots of NIH-funded labs, papers in top journals (or, apparently, “top” journals) — then I despair a bit. Science has gotten so specialized that almost anyone could invent facts or data within their subfield that might pass muster even with close colleagues (even if those colleagues aren’t on the take, he said cynically — there is a mind-boggling amount of money floating around in the drug-development world).

Immediate thought experiment: could Mike or I come up with material for a blog post or paper that would be false but good enough to fool the other? Given how often we find surprising or even counterintuitive results, I think possibly so. I’m not particularly motivated to run the experiment when we’re already digging out from a deep backlog of started-but-never-finished papers, but it remains a morbidly fascinating possibility.

Fang and Casadevall 2011: fig. 1.

Anyway, one problem is that “top” journals have a lot of fraudulent or at least incorrect science in them, roughly corresponding to their impact factors. Now, you might say “yeah but the positive correlation means bad actors get caught”, to which I’d reply “not fast enough” and “how do you know we’re catching all of them?”

Sinking feeling

There’s another problem, I don’t know if it’s equal-and-opposite but it definitely exists: good science that doesn’t float to the top. Here are a couple of quick examples from my neck of the woods:

Working from very little evidence by modern standards, Longman (1933) had correctly figured out that pneumatic sauropod vertebrae come in two flavors, those with a few large chambers and those with many small chambers. He called them “phanerocamerate” and “cryptocamerillan”, corresponding to the independently-derived modern terms “camerate” for the open-chambered form and “camellate” or “somphospondylous” for the honeycombed one. As far as I have been able to determine, nobody paid any attention to this before Wedel (2003b) — Longman’s work on vertebral internal structure wasn’t mentioned or cited by Janensch in the 1940s or Britt or anyone else in the 1990s. To be clear, I’m not putting myself forward as a better researcher than anyone that came before. I just got lucky, to have read a fairly obscure paper while I had my antennae out for any possible mention of pneumaticity.

Speaking of Janensch, his 1947 paper on pneumaticity in dinosaurs was pretty much ignored until the 1990s and early 2000s.

OMNH 1094, a cervical centrum of an apatosaurine, and a crucial player in the Wedel origin story — this was the first vertebra of anything other than Sauroposeidon that Kent Sanders and I scanned.

I owe my career to the Dinosaur Renaissance

Here’s what bothers me about this: I made my career studying pneumaticity in sauropods, buoyed in large part by the fact that I stumbled backwards into a situation where I had access to a big collection of sauropod bones (at the OMNH), free time on a CT scanner (at the university hospital), and a curious and collaborative radiologist (Kent Sanders). But you don’t need a CT scanner to study pneumaticity, as John Fronimos has convincingly demonstrated (see Fronimos 2023 and this post). So why didn’t the revolution in sauropod pneumaticity happen in 1933 or 1947? Or, heck, in 1880 — Seeley and Cope and Marsh and many others recognized that sauropods had highly chambered vertebrae.

I think the most likely explanation is that at the time no-one cared. Pneumatic vertebrae in sauropods were possibly interesting trivia, but sauropods were an evolutionary dead end and so their vertebrae couldn’t tell us anything important about evolutionary success. These attitudes may not have been universal, but they were certainly prevailing.

I had the good fortune to come along at a time when there was renewed interest in dinosaur paleobiology, particularly any characters or body systems shared between non-avian dinosaurs and birds. Suddenly pneumaticity wasn’t some obscure bit of trivia, but the skeletal footprint of a bird-like respiratory system that was potentially a key adaptation for sauropods (Sander et al. 2011) and possibly for dinosaurs more generally (Schachner et al. 2009, 2011). And dinosaurs weren’t any more of an evolutionary dead end than we are, they just happened to mostly not fit into small holes or deep water when the asteroid hit.  (Let’s heat the atmosphere to 400F for a few hours and then make the world dark for a few months or years and then we can talk about evolutionary dead ends.) So adaptations that facilitated dinosaurosity might tell us something about evolutionary success after all.

What are you doing in that cell?

Having a successful career because I happened to hitch a ride on a wave of renewed interest in dinosaur paleobiology is certainly nice, but also worrisome. If it takes 70 or 100 years for the good science to float to the top, does that really count? Whatever convection cells push the good science toward the top would ideally work more like a cook pot on a rapid boil, and not like the imperceptible roiling of Earth’s mantle. So ask yourself: what’s still on its way up to the top right now, that no-one has clocked yet? What’s the Longman (1933) of 2024 — the seemingly incidental observation that is going to seem prophetic in a few decades? Or worse, what was the Longman (1933) of 1994 or 2004, the solid paper that attracted no attention and won’t for another half century?

The convection cell metaphor is particularly apt because a lot of science is siloed. A good idea — say, that the peroneus tertius muscle occurs at a lower frequency in monkeys and apes than in humans, and this tells us something about its evolution — may rise to the top in one cell (comparative anatomy), but not make it over to the neighboring cell (clinical anatomy), where all the happy little molecules think that peroneus tertius is a muscle unique to humans (if you have no idea what I’m on about, see the second numbered point in this post).

So if you want to do good work — in this metaphor, to be at the top where the good science floats (eventually, alongside a seasoning of not-yet-unmasked bad science) — then I think you have to be aware that other cells exist, and occasionally peer into them, if for no other reason than to make sure you don’t accept an idea that’s already been debunked over there. And you need to read broadly and deeply in your own cell — there’s almost certainly valuable stuff you don’t know because the relevant works are stuck to the bottom of the pot. Go knock ’em loose.

References

 

doi:10.59350/27ewm-zn378

The pneumatic rib evolution figure in a more useful format

Here’s a funny thing I hadn’t given much thought to until recently: virtually all journals, even the born-digital variety, have pages in portrait mode for easy printing on 8.5×11 or A4 paper. And many offer a column-width option for figures. So if you want to line up a whole bunch of stuff for easy comparison, for a paper it’s usually easier to orient a figure vertically, like so:

Pneumatic dorsal ribs in a selection of sauropods and their outgroups. King et al. (2024: fig. 3).

And here it is in context on the page:

But virtually all slide presentations use a landscape format, 4:3 for a long time but often 16:9 these days to accommodate wider screens, or phones and tablets in landscape mode. For this a figure much taller than wide is usually not a good use of space, and may present at too small a scale to be readable.

I ran into this last week while prepping a presentation on my research for an anatomy department meeting at work. I wanted to use that King et al. figure because it summed up so much of the paper in one image, but the only version I had was the skyscraper version we’d used in the JVP paper. So I went into GIMP and rotated the image and every element within it by 90 degrees, to produce this landscape version:

I was presenting to an intellectually diverse audience, most of whom do not work on dinosaurs, so I added little silhouettes (my own, cribbed and hacked from all kinds of older work) to make it all more explicable:

This is all my original work, and I’m letting it out in the world here in case anyone else wants to use it. CC-BY like everything else on this blog. FWIW I think mamenchisaurs and diplodocids held their necks elevated — the baseline alert posture for extant tetrapods — I was just moving quickly and more concerned with getting little doodads for all the genera than with any paleobiological implications.

So now I’m wondering if there are any figures in old papers that I’ve avoided putting in talks, possibly subconsciously even, because they’re the wrong shape. Not that I need to do any more navel-gazing than I already do, but maybe something for me to keep an eye out for when I have reason to go back to them (which is often — they’re thought archives).

The more forward-looking takeaway is that if you have to make a taller-than-wide figure to fit a journal page, consider making a wider-than-tall version at the same time to throw into your talks — or vice versa if you’re making the talk first. It’s a time investment for sure, but it may be easier while all the bits are fresh in your head and you have all the elements in separate layers or whatever. Hopefully you already back up the uncompressed versions of all your figures, but Past Matt didn’t always do that, so at least be smarter than that guy!

Tate v2610, a sauropod dorsal rib. Check out the nice deep pneumatic fossa a little way down from the tuberculum of the rib (upper left in the photo).

Parting shot (and an excuse to post a photo for Fossil Friday): on my Tate trip this summer I hit a gang of museums, and everywhere I went I found pneumatic sauropod ribs. I think there are a lot more of these things out there than most folks have appreciated. I’m proud of my recent pneumatic rib papers (Taylor et al. 2023 and King et al. 2024), but I hope they are the just the start of something.

And because I picked that photo: you know what institution has a ton of super-interesting, well-preserved, well-prepped, not-yet-published-on sauropod vertebrae and ribs in a really nicely appointed collections room in an awesome museum run by a small team of excellent human beings? The Tate Geological Museum, that’s who. If you can get yourself to Casper and you have a legit research interest, go check out their collections, there’s SO MUCH good stuff in there. I myself will be back as soon as it can be conveniently arranged.

References

 

doi:10.59350/y1hsw-zvp51

Five questions from Tom Redd

Three years ago, Tom Redd made a very generous commitment to the SV-POW! Patreon, and he remains our most generous donor in total. When I wrote to thank him his reply included “I have thousands of questions about apatosaurus that I would like to ask you some day.”

It seemed only fair to invite him to ask some of those questions, so we asked him to give us five and said we’d try to answer them. When the questions came through, some of them were hard — not really in our area of expertise. But I promised we’d take a crack, and that we’d invite commenters to chip in where we get something wrong or leave something out.

Then a bit more than three years slipped past. Now, finally: here we go!

Question #1:

Do you think sauropods could have evolved long necks as a defensive strategy? (Better to see your enemies from a long way off). The sigmoid curve and the neutral position is another issue!

I recently read a couple of research papers where the researcher used “radiological imaging” to determine the neutral position of the cervical vertebrae. The result was a gently downward curving neck beginning at the pectoral girdle, with the skull only 2 meters off the ground. (I agree with Matt that this would be an ambush predator’s delight!)

Also, could compressive or tensional forces of cartilage and ligaments affect the neutral position of the cervical vertebra? (I believe Matt also alluded to this condition.)

This is the longest question, but maybe the one I’m best positioned to comment on.

First of all, the perennial question of what sauropods’ long necks were for. There are a few candidate explanations out there. The obvious one is that they enabled high browsing, and on the whole we feel that’s the strongest single explanation. Another candidate factor is sexual selection — that sauropods were particularly attracted to long-necked individuals of the opposite sex — but we don’t feel that is a strong explanation for reasons explained in our 2011 paper (Taylor et al. 2011). Predator avoidance would certainly have been aided by the long visual distances allowed by elevated heads, and we are confident that at least some sauropods used their necks in combat — primarily in intraspecific combat (Taylor 2015) but no doubt also against predators when the occasion arose. And we may well have missed other good uses for long necks.

In reality of course all these factors likely played a role: structures do not always, or even often, evolve for a single reason. When people who know much more about ceratopsians tell me “The horns of Triceratops were for intraspecific display and combat”, I don’t doubt them. But I also don’t doubt that, whatever the primary purpose of the horns, a Triceratops confronted by a Tyrannosaurus would do its damnedest to stick its horns into it. In the same way, while high feeding seems like the strongest driver of sauropod neck elongation, the other factors will surely have played in, too.

I’m not sure what radiological imaging papers you have come across, but the one I know about is Berman and Rothschild (2005) in the Thunder Lizards edited volume. This paper rather questionably partitions all sauropod cervicals into two bins, “robust” and “gracile”, and concludes, based on functional stress analysis, that “the robust-type centrum supported a neck held in a vertical, or near-vertical, pose, whereas the gracile-type centrum supported a neck held in a horizontal, or near-horizontal, pose”. If this is right, and their categorization holds, then Camarasaurus and an unidentified titanosaur had vertical necks; and Diplodocus, Apatosaurus, Haplocanthosaurus, Barosaurus and Brachiosaurus all had horizontal necks. We find every part of this unconvincing. At some point we should explain why in detail; but it is not this day.

Finally, yes, compressive and tensile forces in cartilage and ligaments definitely did affect neutral posture. My 2014 paper (Taylor 2014) shows this rather dramatically.

Taylor (2014:figure 3). Effect of adding cartilage to the neutral pose of the neck of Apatosaurus louisae CM 3018. Images of vertebra from Gilmore (1936:plate XXIV). At the bottom, the vertebrae are composed in a horizontal posture. Superimposed, the same vertebrae are shown inclined by the additional extension angles indicated in Table 1. If the slightly sub-horizontal osteological neutral pose of Stevens and Parrish (1999) is correct, then the cartilaginous neutral pose would be correspondingly slightly lower than depicted here, but still much closer to the elevated posture than to horizontal. (Note that the posture shown here would not have been the habitual posture in life: see discussion.)

What we’re seeing here is what the neutral posture would be if cartilage is added to a neck that is otherwise articulated in horizonzal pose. The importance of intervertebral cartilage has often been overlooked, but can make a dramatic difference to neck posture.

Question #2

I recently read a report that indicated Apatosaurus survived at the species level for a period of approximately 8 million years ! So is this a success, average, or a short run?

I’m not sure where you read that, but the problem here is that no-one really knows what Apatosaurus means. We have the type species Apatosaurus ajax, sure, and the referred species Apatosaurus louisae, and the genus Brontosaurus based on the species Brontosaurus excelsus which is sometimes but not always synonymised with the genus Apatosaurus yielding the combination Apatosaurus excelsus, and don’t even get me started on Apatosaurus parvus, Apatosaurus laticollis, Atlantosaurus and whatever the heck AMNH 460 is.

So if we say that Apatosaurus survived for 8 million years, what exactly are we saying? That Apatosaurus ajax is known from sediments that differ in age by 8 million years? That would be interesting if true, but it’s very hard to establish because the referral of any given individual to a particular sauropod species tends to be very uncertain — largely because most specimen are so fragmentary and distorted. And if all we mean is that 8 million years separate the oldest and youngest specimens that have been referred to the genus Apatosaurus — well, that statement is all but meaningless, given the huge uncertainty about what is and is not part of that genus, if indeed genera even really mean anything.

Putting it all together, I’m not confident that there is any reason to think that Apatosaurus was particularly longer lived than other sauropods. Probably Camarasaurus outlasted it if you include all the taxa that have been referred to Camarasaurus. But then I’m far from convinced that that’s the right thing to do, too.

Question #3

Why so many heavy predators during the age of apatosaurs? Predator to prey ratios were in the 6 to 8% range as compared to modern ratios of 2 to 3%!

I’m not sure I can say much about this without knowing the source of the figures, but I assume that what’s being counted here is the number of individuals represented in the fossil record, and the ratio of a predator species to prey species. The problem is that there is a huge amount of vagueness in these numbers, but it’s not obvious that the apatosaur-age figures are comparable to the modern ones.

Consider first the modern ratio. Which animals are counted in each category? In the Serengeti, lions prey on zebras. So far, so simple, but there are also dwarf moongooses, which are predators — but they don’t hunt zebras. So do we count them in the numbers? If so, do we count their prey animals, too? Including invertebrates? And if not, then where do we draw the line between predators that we consider do and do not hunt the prey animals that we’ve decided we’re interested in?

Then there’s the matter of which animals get counted. If you do your Serengeti counts on dead animals, you might find disproportionately many predators because prey animals tend to be consumed. Or you might find disproportionately many prey animals because they tend to die in areas where the corpses are more easily found and counted. You might be able to do better by counting live animals, but then you might easily undercount secretive predators, or perhaps overcount predators because they stand fearlessly around to be counted.

Now consider trying to measure the predator/prey ratio in the Morrison formation. You have all the problems I already mentioned, plus a bunch of others. If you only count complete-ish articulated skeletons then your sample size is too small to be meaningful. If you count isolated elements, you’re at risk of registering multiple instances of the same individual. Counting individuals represented in bonebeds is difficult because of these problems. Assigning an element to a taxon is error-prone (though should generally be OK at the high level of sauropod vs. theropod — or are you?). Bones of different taxa may survive taphonomy better or worse than others. Life history differences will mean that the fossils of long-lived taxa under-represent their live populations. And so on, and on, and on.

Putting it all together, I would tend to be very sceptical that a difference in ratios of 6–8% to 2–3% is necessarily telling us anything.

With all that said, it’s perfectly possible that the average predator:prey body-size ratio was closer in the Morrison than in modern ecosystems. But we’d do better trying to measure that directly from body-fossils than to infer it from population densities.

Question #4

Are all apatosaur tracks on emergent surfaces? (Some depth of water over the prints)

This I don’t know. But then I wouldn’t know how to pick out apatosaur tracks from those of other diplodocids, and I bet no-one else does, either. Tracks are notoriously variable in shape, and can very wildly from the that of the feet that made them. Given that diplodocid feet were mostly pretty similar anyway, I would not be easily persuaded that any track can be confidently identified down to the genus level.

One other thing to be aware of is that there is often not agreement on the conditions under which a given track is made. One palaeontologist may think a given a track is a direct print, another will think it’s an underprint. I don’t mean to say that it’s hopeless and all we can do is throw our hands up in despair — good work is being done on interpreting tracks, but we have a long way to go. And this is not an area that I’m at all expert in.

Question #5

I read recently that in order for a skin impression to be made the Dermal tissue must undergo a type of chemical alteration! Do you think this is what allows the impression to be made?

That doesn’t sound right to me. The first thing that has to happen for an animal to be fossilized is that it needs to be buried in sediment really fast after it dies — before it’s eaten by scavengers. For something as fine as skin impressions to be preserved, that sediment needs to be very fine — which sadly tends to conflict with the first condition, since course sediments can be deposited more quickly than fine ones. It’s really hard for enough fine sediment to be laid down quickly enough to cover an animal of the size of a sauropod, which is why we don’t have sauropod specimens like those gloriously preserved theropods from the Yixian Formation in China(*). So sauropod skin impressions are pretty rare.

(*) Alternatively: there are spectacularly preserved partial sauropod specimens in the Yixian, but Chinese researchers can’t be bothered to write them up because they’d rather spend their time getting a slam-dunk Nature paper out of yet another little feathered theropod. Unduly cynical? Maybe. But I continue to live in hope.

Well, that about wraps up the five questions — to the best of my ability at least. But I’d love to hear from people who know more than I about these topics: leave a comment, and fame and glory could be yours!

And finally … if you, too, would like to have us answer five questions on the sauropod-related topic of your choice, quite possibly in the less than three years, you should consider getting yourself across the The SV-POW! Patreon and making an unreasonably extravagant financial commitment.

References

 

doi:10.59350/3xz8v-8kf38

About that Saurophaganax abstract

The Rebor model of Saurophaganax “Notorious BIG Jungle Variant”, photo from the Big Bad Toy Store.

The SVP 2024 abstract book dropped earlier this week. You can download it here.

I’m not on Twitter, but friends inform me that there’s a lot of discussion going on about the Danison et al. abstract on which I’m an author (pp. 164-165 in the abstract book, or I excerpted it as a separate file at this link).

Left: unpainted cast of the thumb claw of the giant Oklahoma allosaurid (referred to Saurophaganax, as of this writing and for at least a little while longer). Right: a commercially available cast of a big Allosaurus thumb claw (dunno which one, I bought this in the Dinosaur National Monument gift shop a few years ago, maybe some CLDQ specimen?). Old school wooden ruler for scale.

I’m not going to give any more technical information right now than you can read in the abstract. But if you want more, I have two pieces of good news. First, if you’re going to SVP you can catch Andy Danison’s talk on Friday morning in the Theropoda 1 session.

Second, and of more global reach, we — the same author team from the abstract — have a long, lavishly illustrated paper in revision at an open access journal. The paper goes into far more detail, and with all the evidence we could bring to bear, on the identity of the type and referred material of Saurophaganax. And it will be freely available to the world once it’s published.

This tweet is flattering (misspelled name notwithstanding). It hits on something important, though: all of us on the author slate knew this was a big swing, and that it would be heavily scrutinized by both professional paleontologists and folks with an avocational interest, and we wouldn’t have put our names to it if we weren’t pretty darned sure we were right. That doesn’t mean our hypothesis is any more correct! Science runs on evidence, not reputations (or at least it should). But we are all putting our reputations on the line here to some extent, which at least tells you what we think of the evidence. So, yeah, as the abstract says, in our estimation the axial elements used to diagnose Saurophaganax, including the holotype, belong to one or more sauropods. (If you disagree, that’s cool. I’m not going to engage in any skirmishes right now when our battleship will be in range shortly.)

That still leaves a big pile of material from a really big allosaurid. Obviously any taxonomic acts coming out of this project will be in the paper, and not in the talk or the abstract. With any luck, we won’t have long to wait.

And, like Forrest Gump, that’s all I have to say about that.

Paramedullary diverticula in ostriches. If you like this sort of thing…hold onto your butts! Atterholt and Wedel (2022: fig. 3).

Since I don’t know if I’ll get around to posting again before SVP, I should shout out a couple more presentations. One is the Atterholt et al. talk on paramedullary diverticula in extinct ornithodirans, including pterosaurs, non-avian dinosaurs, and fossil birds — that’s the other one on which I’m an author. Abstract on page 85 of the abstract book, and Jessie’s talk will be on Friday afternoon. Another is the Hart et al. poster on a vertebra of the tomistomine croc Thecachampsa. Abstract on page 243, poster will be Saturday afternoon. I’m not an author on this one, not involved in the research, but if you read the abstract in light of recent activity here, you’ll know why I’m stoked about it!

References

  • Atterholt, J., Wedel, M., Benito, J., and Field, D.J. 2024. Evidence of paramedullary diverticula in extinct ornithodirans. Society of Vertebrate Paleontology 84th Annual Meeting Program, p. 85.
  • Danison, A., Woodward, H.N., Barta, D.E., Wedel, M., Lee, A.H., Flora, H., and Snively, E. 2024. Osteohistology, probable chimerism, and taxonomic revision of Saurophaganax maximus. Society of Vertebrate Paleontology 84th Annual Meeting Program, pp. 164-165.
  • Hart, W.J., Hill, R.V., and Bennington, J.B. 2024. A caudal vertebra from Thecachampsa sp. (Crocodylia: Tomistominae) with comments on systematics. Society of Vertebrate Paleontology 84th Annual Meeting Program, p. 243.

 

doi:10.59350/10vtq-4tb98

More “artificial intelligence” idiocy

I asked ChatGPT a very simple question:

A man needs to cross a river. He has a goat with him, which he needs to take across. The boat is big enough to carry the man and only one other object. How should he do it?

Here is the “solution” it suggested:

This is a classic river-crossing puzzle! The man needs to cross the river with a goat, but since the boat can only carry the man and one other object at a time, he must plan his crossings carefully. Here’s how he should do it:

  1. Take the goat across the river first and leave it on the other side.
  2. Return alone to the original side.
  3. Take an empty trip across the river.
  4. Take the

That’s it. That’s the whole thing. Here’s the screenshot if you don’t believe me:

Screenshot of ChatGPT answering my question.

This is a particularly lovely demonstration of the emptiness of LLMs. It recognises the words from the classic river-crossing puzzle, and generates a sequence of words resembling those to a solution.

But they are entirely devoid of meaning.

I don’t just mean that they happen to be incorrect in the present case, where the correct solution is of course just “take the goat across the river” and you’re done. I mean that if you ask the classic version of the question, with the wolf and the sack of grain, and if the LLM happens to emit a sequence of words that resembles a correct answer, those words are still devoid of meaning.

It’s really important that we get this. LLMs do not reason. At all.

 

doi:10.59350/1sf2n-p3s29